JP6546828B2 - Modify at least one parameter used in video processing - Google Patents

Description

  The invention relates to the field of video surveillance. In particular, it relates to a method, apparatus and system for modifying at least one parameter used by a video processing algorithm for on-site monitoring.

  In many camera surveillance situations, it is important to detect events such as the presence or movement of objects in the distance. Of course, objects located far away from the camera will be small in the image captured by the camera. For example, they may only contain a few pixels in the image. Thus, in order to detect distant events, it is necessary, for example, to be able to detect small appearing objects in the captured image.

  However, detecting small objects usually comes at the cost of increased false positives. This relates to the fact that noise is present in the image taken by the camera. Noise can easily be mistaken for small objects, and vice versa, thereby leading to false positives. To reduce false positives, video processing algorithms for object detection, motion detection, and / or object tracking include morphological filters to remove small detections, brightness change thresholds and noise filtering etc. Including the parameters of The parameters are usually set such that false detections caused by noise are eliminated but there is still sufficient sensitivity to detect real events. In other words, there is a tradeoff between detection range and avoiding false positives.

  The problem of detecting and identifying distant objects has been recognized. In order to solve this problem, it has been proposed to use a thermal camera to identify the position coordinates of a person. The position coordinates are then input to a second camera for capturing the magnified image of the person, and in particular to control means for controlling its direction and zoom.

  Such a method has the disadvantage that the second camera is controlled to change its direction and zoom level to follow a particular target object. Thus, such a method is only suitable for detecting one object at a time.

  Therefore, there is room for improvement. In particular, there is a need for alternative methods that allow for a wide range of detection while reducing false positives.

  Accordingly, in view of the above, it is an object of the present invention to provide a method and apparatus that enables detection of a wide range of objects and / or activity while reducing false positives.

According to a first aspect of the present invention, the above object is to correct at least one parameter used by a video processing algorithm for scene monitoring, such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm. Achieved by the method, the method
Receiving a first video sequence and a second video sequence of the scene, wherein the first video sequence is imaged with a thermal camera so as to include thermal information indicative of the temperature of the scene When,
Applying a video processing algorithm to the second video sequence, wherein at least one parameter used by the video processing algorithm is modified to be temperature dependent based on the thermal information contained in the first video sequence Applied, including.

  According to the method, thermal information in a first video sequence imaged by a thermal camera is used to modify or change at least one parameter of a video processing algorithm applied to a second video sequence. Thus, at least one parameter is made dependent on the temperature at the site. For example, modifying at least one parameter so that the video processing algorithm is sensitive to small detections in selective areas with temperatures of interest, such as human body temperature, and not in other areas. Can be adapted. As a result, in the area corresponding to the temperature of interest, a wider detection range is possible, while in the areas corresponding to the other temperature ranges false detections are suppressed low.

  A video processing algorithm is an image processing algorithm applied to a sequence of images, but may be an algorithm for the detection of an event in the field. This includes algorithms for object detection, motion detection, and object tracking. It should be noted that the video processing algorithm is not an algorithm that controls the functions of the camera itself.

  It is generally meant that by modifying at least one parameter, the value of the parameter is modified, ie changed. For example, the value of the parameter may be increased or decreased.

  In general, at least one parameter may be modified with respect to the entire image of the second video sequence. For example, if the thermal information in the first video sequence indicates that an object having a temperature that is at a temperature of interest is present in the field, then the video processing algorithm is small over the image plane of the second video sequence image. At least one parameter may be altered to be sensitive to detection.

  At least one parameter may also be corrected throughout the image, but as a function of temperature. That is, nothing happens to the numerical value of the parameter in the non-interest temperature region, and the parameter becomes a new value in the temperature region of interest.

  In particular, at least one parameter used by the video processing algorithm is modified with respect to the area in the image of the second video sequence based on thermal information of the corresponding area in the image of the first video sequence It can be done. Thus, the video processing algorithm is made more sensitive to small detections in some areas that may have thermal information of interest, and less sensitive to small detections in other areas (hence, Low sensitivity to noise).

  Thermal information generally means any information indicative of the temperature at the site. For example, the thermal information may include the brightness values of the images in the first video sequence indicating the temperature of the scene. Thus, at least one parameter used by the video processing algorithm may be modified based on the luminance values of the images in the first video sequence.

  Video processing algorithms may have the purpose of detecting different types of events in the scene, such as the presence or movement of a particular type of object or several different types of objects. Different types of objects may be distinguished based on thermal information in the first video sequence as they may be associated with different temperature ranges. For example, a human body may be associated with one temperature range, while an operating engine of a car may be associated with another temperature range. Thus, depending on the specific purpose of the video processing algorithm, there may be at least one predetermined temperature range of interest that may be used to distinguish different regions in the image of the first video sequence. The at least one predetermined temperature range of interest may include a temperature range corresponding to human body temperature. Such predetermined temperature range may, for example, be specified by the user.

This method is
Identifying, in the image of the first video sequence taken by the thermal camera, a pixel having a luminance value indicative of a temperature lying in at least one predetermined temperature range of interest;
At least one parameter used by the video processing algorithm may include being modified with respect to a region of the image of the second video sequence corresponding to a pixel identified in the image of the first video sequence.

  In other words, pixel regions having thermal information of interest are identified in the first video sequence, and at least one parameter is corrected for the corresponding region in the second video sequence. Thus, at least one parameter is modified in an area having certain temperature information of interest, such as an area where a human is present, and a video processing algorithm for detecting an event in such area High sensitivity.

  As should be understood, the first video sequence and the second video sequence may have different resolutions and may be imaged from slightly different viewing angles of the scene. However, there is still a correspondence between the first video sequence and the second video sequence. More specifically, if the pixels or areas in the image of the first video sequence and the pixels or areas in the image of the second video sequence represent the same point in the scene, then the image of the first video sequence The pixels or regions in correspond to the pixels or regions in the image of the second video sequence.

  In accordance with the above, at least one parameter may be corrected for various regions in the image of the second video sequence, which regions are identified based on the thermal information in the first video sequence. Thus, at least one parameter may differ depending on the image of the second video sequence, which difference is determined by the thermal information in the first video sequence. In particular, at least one parameter used by the video processing algorithm may be modified based on the luminance value of the image of the first video sequence, the at least one parameter being the luminance value of the image of the first video sequence In response, it will change from image to image of the second video sequence. The difference may be, for example, continuous or stepwise. This is advantageous in that at least one parameter can be adapted depending on the temperature in the first video sequence.

  The thermal information in the first video sequence is not only to decide where, i.e. in which region at least one parameter should be corrected, but also to decide how or how at least one parameter should be corrected. It should be noted that it can also be used. By allowing at least one parameter to differ from the image of the second video sequence as a function of the luminance value of the image of the first video sequence, the magnitude of the correction is also determined by the luminance of the image of the first video sequence It is a function of the value. For example, the amount of increase or decrease of the parameter may be a function of the luminance value of the image of the first video sequence.

  At least one parameter may be a parameter associated with the detection of an event in the field. This includes parameters for object detection, motion detection, and object tracking.

  For example, at least one parameter may include a threshold for detection of an event, such as motion or an object present. In particular, the video processing algorithm may include comparing the difference in luminance values between subsequent images in the second video sequence to a threshold. The threshold is corrected based on the thermal information in the first video sequence. For example, the threshold may be increased in the area corresponding to the thermal information of interest. In this way, the video processing algorithm can be made sensitive (also sensitive to noise) to the detection of events in the area of interest and low in other areas.

  At least one parameter may include a parameter used in a filter for the removal of small detections (which may be noise), such as the removal of adjacent or connected small groups of pixels, also called pixel clusters. For example, the at least one parameter may include a threshold for the size of the neighboring pixels such that neighboring pixels having a size below the threshold are removed. In this way, the video processing algorithm can be made sensitive (also sensitive to noise) to the detection of events in the area of interest and low in other areas.

  According to an example, the video processing algorithm comprises smoothing filtering, wherein at least one parameter of said smoothing filtering is modified based on the thermal information in the first video sequence. By smoothing filtering is meant applying a smoothing filter. Smoothing filtering may be applied to the images of the second video sequence. The smoothing filter may be a temporal filter (with a smoothing effect in the temporal direction) and / or a spatial filter (with a smoothing effect in space). Examples of spatial smoothing filters are morphological filters and median filters.

  Thus, smoothing filtering may include morphological filtering for the removal of small detections (which may be noise). In particular, the smoothing filtering may include morphological filtering of the image of the second video sequence using a structuring element, the structuring element being modified based on thermal information in the first video sequence . The modification may include modifying at least one of the size and the shape of the structuring element. In this way, the video processing algorithm can be made sensitive (also sensitive to noise) to the detection of events in the area of interest and low in other areas.

  The thermal camera may include a far infrared (LWIR) sensor. LWIR sensors are advantageous in that they complement visible light sensors in a way that other infrared sensors can not do.

  A second video sequence may be imaged by a visible light camera. This is advantageous in that visible light cameras usually have higher resolution than, for example, thermal cameras, thereby simplifying detection and tracking.

  A second video sequence may be taken by the thermal camera. This is advantageous, for example, in that thermal cameras do not require external light to capture useful images at night and other low light conditions. Another advantage is that only one camera needs to be provided.

  According to a second aspect of the present invention, the above object is achieved by a computer program product comprising a computer readable storage medium having instructions adapted to perform the method of the first aspect when executed by a processor To be achieved.

According to a third aspect of the invention, the above object is to correct at least one parameter used by a video processing algorithm for scene monitoring, such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm. Achieved by the processing device, the device
A receiver configured to receive a first video sequence and a second video sequence of the scene, wherein the first video sequence is imaged using a thermal camera such that the first video sequence includes thermal information indicative of the temperature of the scene The receiver to be
A video processing component configured to apply a video processing algorithm to a second video sequence to monitor a scene, for use by the video processing algorithm based on thermal information included in the first video sequence And video processing components configured to modify at least one of the parameters being performed, such that at least one parameter of the video processing algorithm is temperature dependent.

According to a fourth aspect of the invention, the above object is to correct at least one parameter used by a video processing algorithm for scene monitoring, such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm. Achieved by the system, the system
At least one camera configured to image the first video sequence and the second video sequence in the scene, the thermal camera adapted to image the first video sequence, such that At least one camera, such that the first video sequence includes thermal information indicative of the temperature at the site;
A processing unit according to a third aspect configured to receive and process the first video sequence and the second video sequence and to modify parameters used by the video processing algorithm.

  The second, third and fourth aspects may generally have the same features and advantages as the first aspect. It is further noted that the invention relates to all possible combinations of features, unless explicitly stated otherwise.

  In addition to the above, further objects, features, and advantages of the present invention will be better through reference to the accompanying drawings and through the following illustrative and non-limiting detailed description of preferred embodiments of the present invention. It will be understood. The same reference numbers are used for similar elements.

FIG. 8 illustrates a system for modifying at least one parameter used by a video processing algorithm for on-site monitoring, according to an embodiment. FIG. 5 is a flow chart of a method for modifying at least one parameter used by a video processing algorithm for site surveillance according to an embodiment. Fig. 6 illustrates the modification of parameters to be used by the video processing algorithm based on thermal information according to an embodiment. Fig. 6 shows an example of object detection where the parameters are not corrected prior to application of the object detection algorithm. Fig. 6 shows an example of object detection where the parameters are modified according to an embodiment prior to application of the object detection algorithm.

  The invention will now be described in more detail with reference to the accompanying drawings in which embodiments of the invention are shown. The systems and apparatus disclosed herein are described in operation.

  FIG. 1 shows a system 100 for monitoring a site 102. The system 100 includes at least one camera 104a and 104b and a processing unit 106, here represented by two cameras.

  Site 102 may include a number of objects 102a-f, in this case illustrated by person 102a, person 102b, car 102c, car 102d, road 102e, and tree 102f. Some of the objects may be moving, such as person 102a, person 102b, car 102c, and car 102d, and other objects such as road 102e and tree 102f (at least when the wind is not blowing) It may be stationary. The objects 102a-f may also be of various sizes and may be located at different distances from the cameras 104a-b. For example, objects 102a and 102c are located further from camera 104a-b as compared to objects 102b, 102d and 102f. Objects that are small and / or remote from the cameras 104a-b may be difficult to distinguish among the images captured by at least some of the cameras 104a-b. Objects 102a-f may also be associated with different temperatures or temperature ranges. For example, persons 102a-b may be associated with a temperature range corresponding to human body temperature, and cars 102c-d may be associated with a temperature range corresponding to the temperature of the operating engine. Other objects, such as trees 102f, do not generate heat themselves and are usually associated with a temperature range corresponding to the ambient temperature.

  One of the at least one camera 104a-b, here the camera 104a, is a thermal camera. Preferably, the camera 104a includes a far infrared (LWIR) sensor. LWIR sensors are advantageous in that they complement the visible light camera differently than other infrared sensors. Another camera of the at least one camera 104a-b, such as the camera 104b, may be a visible light camera. According to another example, the second camera 104b may also be a thermal camera. The cameras 104a-b are shown as separate cameras, mainly for illustration reasons. However, it should be understood that the at least one camera 104a-b may be combined into a single unit, such as a combined thermal and visible light camera.

  At least one camera 104a-b is arranged to view the scene 102 and to capture a video sequence of the scene 102 for surveillance purposes. More particularly, at least one camera 104a-b is arranged to image the first video sequence 105a and the second video sequence 105b of the scene 102. If the system 100 includes only one camera 104a (which in this case is a thermal camera), the first video sequence 105a and the second video sequence 105b may be imaged by the same camera. The first video sequence 105a and the second video sequence 105b may even be the same video sequence. However, in the example shown, the first video sequence 105a is imaged by the first (thermal) camera 104a and the second video sequence 105b is imaged by the second camera 104b.

  The processing unit 106 includes a receiver 108 and a video processing component 110 in communication with the receiver 108. The processing device 106 may be provided separately from the at least one camera 104a-b or may be included in one of the cameras 104a-b. In any case, the processing unit 106 is operatively connected to the at least one camera 104a-b in a wired or wireless connection and via the receiver 108 from the at least one camera 104a-b to the first video sequence 105a and a second video sequence 105b may be received.

  The video processing component 110 is configured to apply video processing algorithms to the second video sequence 105b for monitoring of the scene 102. For example, video processing component 110 may be configured to detect an event, such as the presence and / or movement of an object at scene 102, as well as perform object tracking. The video processing algorithm applied for that purpose may be, for example, any known algorithm for motion detection, object detection, and / or object tracking.

  Video processing algorithms of the type described above are typically associated with at least one parameter, including parameters associated with the detection of events, such as thresholds and parameters associated with various filters. Video processing component 110 is configured to modify at least one parameter prior to application of the video processing algorithm. As described in more detail below, such a modification of at least one parameter is based on the information in the first video sequence 105a.

  Video processing component 110 may be implemented in hardware, software, or a combination thereof. In particular, video processing component 110 may include memory 112 and processor 114, such as a central processing unit. Memory 112 may serve as a (non-transitory) computer readable medium or storage device, and may store computer code instructions for any of the methods described herein. In particular, computer code instructions are adapted to perform any of the methods described herein when executed by processor 114.

  The operation of system 100 and, in particular, the operation of processing device 106 will now be described with reference to the flowcharts of FIGS. 1, 3 and 2.

  In step S02, the processing device 106 receives the first video sequence 105a and the second video sequence 105b from the at least one camera 104a-b via the receiver 108. In the following example, it is assumed that the first video sequence 105a is from the thermal camera 104a and the second video sequence 105b is from the visible light camera 104b.

  The first video sequence 105a captured by the thermal camera 104a, and in particular, the images in the first video sequence 105a, include thermal information indicative of the temperature at the scene. The thermal information may include, for example, luminance values of the images in the first video sequence 105a, the luminance values indicating the temperature at the scene 102. The temperature may be an absolute temperature or may only represent some temperature deviation from the rest of the site. FIG. 3 schematically shows an image 305a of the scene 102 in a first video sequence 105a. In the image 305a, the persons 102a-b are shown with the luminance corresponding to the first temperature range which is the temperature range corresponding to the human body temperature. In image 305a this is shown by the person having the first pattern. Similarly, the car 102c and the car 102d are displayed at the brightness corresponding to the second temperature range which is a temperature range corresponding to, for example, the temperature of the operating engine. In image 305b this is shown by a car having a second pattern.

  Continuing, at step S04, the video processing component 110 of the processing unit 108 modifies at least one parameter of the video processing algorithm to be applied to the second video sequence 105b.

  By default (as known in the prior art), the video processing algorithm will apply the same value of at least one parameter across the image planes of the images in the second video sequence 105b. This is further illustrated in FIG. 3 which shows an image 307a from the second video sequence 105b. In the image 307a, the values of the parameters 309 of the video processing algorithm are superimposed as indicated by the dot pattern. The values of the parameters are the same across the image plane of the image 307a.

  Video processing component 110 modifies at least one parameter 309 to cause parameter 309 to be dependent on the temperature at site 102. For this purpose, the video processing unit 110 corrects the parameters 309 based on the thermal information contained in the first video sequence 105a, as shown for example by the image 305a. For example, video processing unit 110 may modify parameter 309 to take updated values across the image plane of the image of second video sequence 105b.

  More preferably, the parameters 309 may be modified to take different values in different parts or regions of the image 307a of the second video sequence 105b. For example, as shown in image 307b of FIG. 3, parameter 309 takes a first value 309a in the first region, a second value 309b in the second region, and a third in the third region. Can take the value 309c.

  Regions may be found by considering thermal information in the first video sequence, such as in image 305a. More specifically, the regions in the image 307b in the second video sequence 105b may correspond to some regions in the image 305a of the first video sequence 105a. Here, it is assumed that the image 305a and the image 307b are temporally aligned. Regions in the image 305a may be identified based on the thermal information. For example, the region may be a pixel having a luminance value (ie, a temperature value) corresponding to a predetermined temperature range of interest, such as a temperature range corresponding to human body temperature or a temperature range corresponding to the temperature of the operating engine. Can be identified in the image 305a. By identifying the area as such, the first area can be identified as a pixel corresponding to person 102a, person 102b, and the second area can be identified as a pixel corresponding to car 102c, car 102d. Thus, video processing component 110 may modify parameter 309 to take different values in different regions. In the example shown, therefore, the parameter takes the first value 309a in the area where it can be expected to find humans and the second value 309b in the area which can be expected to find the car with the engine in operation. Will be corrected to take.

  The thermal information in the first video sequence 105a can not only be used to decide where in the image plane the parameter 309 should be modified. It can also be used to determine the magnitude of the correction, ie how much the parameter 309 should be corrected or adjusted. Typically, the magnitude of the correction may be set as a function of the thermal information in the first video sequence 105a. More specifically, luminance values in the image of the first video sequence 105a, such as the image 305a, indicate the temperature. Thus, the magnitude of the parameter correction may be set as a function of the brightness of the image 305a of the first video sequence 105a. In the example shown, therefore, the first value 309a may be set according to the luminance value of the pixel corresponding to the person 102a, the person 102b in the image 305a, and the second value 309b is in the image 305a. It can be set according to the luminance value of the pixel corresponding to the car 102c and the car 102d.

  In the above example, it has been described that the parameters 309 can take different values in different parts or regions of the image plane. This does not necessarily mean that portions or regions of the image plane have strict boundaries. More generally, parameters 309 may be modified differently in image 307b as a function of the luminance value of the corresponding pixel in image 305a. As such, since the luminance values (temperatures) in the image 305a vary spatially within the image plane of the image 305a, the value of the parameter 309 will vary spatially within the image plane of the image 307b. I will.

  At least one parameter 309 may be, for example, a threshold used to detect motion, detect an object, and / or track an object. More specifically, it may be a threshold for detecting a change in luminance value between successive images in the second video sequence 105b. For example, the video processing algorithm may compare the difference in luminance values between subsequent images in the second video sequence 105b to a threshold to detect an event such as motion at the scene 102. The luminance change threshold may be modified by the video processing component 110 at step S04 prior to application of the video processing algorithm. For example, the threshold of brightness change is lowered in the portion or region 309a and / or 309b of the image 307b at which the temperature provided by the image 305a of the thermal camera 104a is found to be within the temperature range of interest. Can be Thus, the video processing algorithm will be sensitive to changes in luminance in those portions or areas of the image 307b of the second video sequence 105b. In other areas, the brightness change threshold may for example be raised or left uncorrected. In this way, false positives caused by events in other temperature ranges can be suppressed. Further, as noted above, video processing component 110 may also modify, ie, raise or lower, the threshold of brightness change as a function of the brightness in image 305a of first video sequence 105a.

  At least one parameter may be associated with noise filtering and a filter for removal of small detections (possibly caused by noise). For example, there may be parameters that set a threshold for the size of the neighboring pixel group. It may be determined that an object or motion is present if the video processing component 110 detects in the images of the second video sequence, or in the differences between the images, neighboring pixels above the threshold. If the size is below the threshold, it may be determined that neighboring pixels are too small to be removed to avoid false positives.

  Video processing component 110 may also remove detection of small objects and / or motion by applying morphological filters, such as filters for erosion, to the images or differences between images of second video sequence 105b. In particular, the video processing component 110 may morphologically filter the images or the differences between the images of the second video sequence 105b by using structuring elements. Video processing component 110 may modify the structuring element in step S04 prior to application of the video processing algorithm. For example, video processing component 110 may modify the size and / or shape of the structuring element. For example, the video processing component 110 may, for example, reduce the size of the structuring element used for erosion, etc., in the part or the area of the image 307b of the second video sequence 105b, corresponding to the temperature range of interest. In parts or areas, the size of the structuring element may be not increased or corrected. Thus, the morphological filtering operation leaves more (and less) detection in the area corresponding to the temperature range of interest, as compared to the other areas, whereby the video processing algorithm is of interest. The sensitivity to detection may be increased in the region corresponding to a temperature range.

  The video processing algorithm may be a motion detection algorithm. From a simple comparison of two successive images and counting the number of pixels with varying luminance values between the two images, for example, ranging from more complex algorithms using spatio-temporal slicing Various motion detection algorithms are known. In a simple pixel difference algorithm, a threshold for luminance value differences may be used to reduce false positives, and the luminance value differences must exceed the threshold to be considered as changes. Thermal information may be used to correct such thresholds. Furthermore, the number of changed pixels may be compared to a predetermined reference number, and a number of changed pixels below the reference number will not be considered motion, but if more pixels than the reference number are changed , Would be considered a move. Thermal information may be used to correct such reference numbers.

  More sophisticated motion detection algorithms may involve forming a background model of several previous images and comparing a new image to the background model, rather than comparing one image to another. The background model may be formed by infinite impulse response filtering. In this way, temporal filtering may be achieved to reduce false detection caused by noise. In addition to the parameters already discussed for the simple motion detection algorithm, thermal information is used in such a temporal filtering method, for example, to correct the influence of each of the new images, ie the new image is the background The coefficients that are weighted into the model may be modified.

  In US Pat. No. 8,121,424, a motion detection algorithm is disclosed which uses spatio-temporal slicing. Here, the images of the video sequence are combined into an image volume having vertices X, Y and t. Here, X and Y are space coordinates, and t represents time. The image volume is sliced in either the (X, t) or (Y, t) plane. A search is performed along the search line in the video sequence to locate the motion. Motion can be detected by identifying lines that are not parallel to any of the vertices. Stationary objects and illumination changes will instead appear as lines parallel to one of the vertices. Here, thermal information needs to be detected, for example, to correct the period of time t in which the image in the image volume is taken, to correct the spatial size of the image volume, and / or as motion. It may be used to correct the threshold for the number of adjacent similar pixels along the search line.

  In step S06, the video processing component 110 receives the second video sequence 105b with a video processing algorithm using at least one modified parameter 309a, 309b, 309c. In general, the video processing algorithm may be any type of video processing algorithm that provides an output that may be used when monitoring or monitoring the scene 102. In particular, the video processing algorithm may be an algorithm for detecting distant events, such as in perimeter surveillance. This may include motion detection algorithms, object detection algorithms, and / or object tracking algorithms.

  An example will now be described with reference to FIGS. 4a and 4b to illustrate some of the benefits of the disclosed method. In this example, the video processing algorithm is an algorithm for object detection, which is used by the video processing component 110 to determine whether an object is present, eg, the number of neighboring pixels It is assumed that it is associated with a parameter that is a threshold for the size of the object at point.

  Fig. 4a shows the case where the parameters are not modified before application of the video processing algorithm to the images of the second video sequence 105b, and Fig. 4b video processing to the images of the second video sequence 105b according to an embodiment. We show the case where the parameters are modified before applying the algorithm.

  In FIG. 4a, the parameter 409 takes the same value across the image plane of the image 407a of the second video sequence 105b. The value of parameter 409 is usually such that possible object detection is eliminated as non-objective, so that the number of erroneous object detections is reduced, ie, some pixels below parameter 409 are removed. It is set. However, this sacrifices the possibility of detecting small objects, such as objects located far from the camera. This is further shown in an image 411a showing the objects identified in the image 407a by the video processing component 110. In this case, the video processing algorithm managed to detect objects 102b, 102d and 102f. However, among other things, objects 102a and 102c located far from camera 104b in scene 102 have not been detected. This is because their magnitude is below the threshold of parameter 409.

  In FIG. 4b, the parameters are modified by the video processing component 110 prior to application of the object detection algorithm (see step S04 above). As explained with reference to FIG. 3, the parameters are therefore different values 409a, 409 in different areas of the image 407b of the second video sequence 105b, depending on the thermal information in the first video sequence 105a. , Corrected to take 409c. For example, the parameters may be set to take lower values 409a, 409b in the region corresponding to the temperature range of interest of the image 407b as compared to the other regions of the image 407b. In this way, smaller objects can be detected in the area corresponding to the temperature range of interest compared to the other areas. This is further shown in an image 411b showing the object identified in the image 407b by the video processing component 110. In this case, the video processing algorithm somehow detected the objects 102a and 102c with the modified values 409a, 409b of the object detection threshold in these areas.

  It will be appreciated that one of ordinary skill in the art can modify the above-described embodiments in many ways and still use the advantages of the invention as set forth in the above-described embodiments. Accordingly, the present invention should not be limited to the illustrated embodiments, but should be defined only by the appended claims. Further, the illustrated embodiments may be combined as understood by one of ordinary skill in the art.

Claims (14)

A method of modifying at least one parameter (309, 409) used by a video processing algorithm for surveillance of a scene (102), such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm, comprising:
Receiving (S02) a first video sequence (105a) and a second video sequence (105b) of the scene (102), the first video sequence (105a) being of the scene (102) Imaged, received using a thermal camera (104a) to include thermal information representative of temperature;
Correcting (S04) the numerical value of at least one parameter (309, 409) used by the video processing algorithm based on the thermal information included in the first video sequence (105a);
The at least one parameter (309, 409) of the video processing algorithm becomes temperature dependent;
It said thermal information is numerically the at least one parameter in which region of the second video sequence is modified, and, to decide value of said at least one parameter in the region how much is modified It is used to,
The numerical value of the at least one parameter remains uncorrected outside the region ,
To fix and
Applying (S06) the video processing algorithm to the second video sequence (105b). The numerical value of the at least one parameter (309, 409) used by the video processing algorithm is the first video sequence (105a) with respect to the region in the image (307b, 407b) of the second video sequence (105b). The method according to claim 1, wherein the correction is made based on the thermal information of the corresponding area (102a, 102b, 102c, 102d) in the image (305a) of. The thermal information includes luminance values of the image (305a) in the first video sequence (105a), the luminance value representing the temperature of the scene (102), the at least one used by the video processing algorithm The method according to claim 1 or 2, wherein the numerical value of one parameter (309, 409) is modified based on the luminance value of the image (305a) in the first video sequence (105a). In the image (305a) of the first video sequence (105a) taken by the thermal camera (104a), a pixel (102a,) having a luminance value indicative of a temperature lying in at least one predetermined temperature range of interest Further comprising identifying 102b, 102c, 102d),
The pixels (102a, 102b, 102c, 102d) identified in the image (305a) of the first video sequence (105a) are numerical values of the at least one parameter (309, 409) used by the video processing algorithm. The method according to claim 3, wherein the region of the image (307b, 407b) of the second video sequence (105b) corresponding to.   5. The method of claim 4, wherein the at least one predetermined temperature range of interest comprises a temperature range corresponding to human body temperature.   The video processing algorithm comprises comparing the difference in luminance value between successive images in the second video sequence (105b) to a threshold, the threshold being the first video sequence (105a). The method according to any one of claims 1 to 5, wherein the method is corrected based on the thermal information in the medium.   The video processing algorithm comprises smoothing filtering, wherein at least one parameter of the smoothing filtering is modified based on thermal information in the first video sequence (105a). Method described in Section.   The smoothing filtering comprises morphological filtering of the image (307b, 407b) of the second video sequence (105b), wherein the morphological filtering uses a structuring element, the structuring element being the first The method according to claim 7, wherein the method is modified based on thermal information in the video sequence (105a).   The method according to any of the preceding claims, wherein the thermal camera (104a) comprises a far infrared (LWIR) sensor.   The method according to any of the preceding claims, wherein the second video sequence (105b) is imaged with a visible light camera.   The method according to any of the preceding claims, wherein the second video sequence (105b) is imaged with the thermal camera (104a).   A computer program product comprising a computer readable storage medium (112) having instructions which, when executed by a processor (114), perform the method according to any one of the preceding claims. A processing unit (106) for modifying at least one parameter (309, 409) used by a video processing algorithm for surveillance of a scene (102), such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm.
A receiver (108) for receiving the first video sequence (105a) and the second video sequence (105b) of the scene (102), the first video sequence (105a) comprising a thermal camera (104a) A receiver, which is imaged using the thermal information representing the temperature of the site (102);
A video processing component (110),
Based on the thermal information contained in the first video sequence (105a), modify the numerical value of at least one parameter (309, 409) used by the video processing algorithm and monitor the scene (102) A video processing component (110) applying the video processing algorithm to the second video sequence (105b) to
Making the at least one parameter of the video processing algorithm temperature dependent as a result of the correction;
It said thermal information is numerically the at least one parameter in which region of the second video sequence is modified, and, to decide value of said at least one parameter in the region how much is modified And a video processing component used in the at least one parameter value remains uncorrected outside the region . A system (100) for modifying at least one parameter (309, 409) used by a video processing algorithm for surveillance of a scene (102), such as a motion detection algorithm, an object detection algorithm, or an object tracking algorithm.
At least one camera (104a, 104b) for imaging a first video sequence (105a) and a second video sequence (105b) of said scene (102), said first video sequence (105a) being said At least one camera including a thermal camera (104a) imaging the first video sequence (105a) to include thermal information representative of the temperature of the scene (102);
14. The processing unit (106) according to claim 13, receiving and processing the first video sequence (105a) and the second video sequence (105b) and modifying parameters used by the video processing algorithm. Including systems.

Images